THE INTERACTION BETWEEN NUMERALS AND NOUNS James R Hurford 1 Introduction Aims This paper is a descriptive survey of the principal phenomena surrounding cardinal numerals in attribution to nouns, with some concentration on European languages, but within a world-wide perspective. The paper is focussed on de- scribing the syntagmatic distribution and the internal structure of numerals. By contrast, the important topic of the paradigmatic context of numerals, that is how their structure and behavior relates to those of quantifiers, determiners, adjectives, and nouns, does not receive systematic discussion here, although many relevant comments are made in passing. A further necessary limitation in scope is the exclusion of forms which are only marginally cardinal numerals, if at all, such as English both, dozen, fourscore, pair, triple and their counterparts in other lan- guages. Organization. This paper is organized into successive major sections as fol- lows: 1. Simple lexical numerals (Section 2). 2. Simple lexical numerals modifying nouns (Section 3). 3. Complex numerals: internal structure (Section 4). 4. Complex numerals modifying nouns (Section 5). This organization mirrors a diachronic hypothesis about the emergence of the phenomena in languages. According to this hypothesis, some (but not all) internal structural properties of complex numerals are patterned after the ways in which 1 simple lexical numerals modify nouns. Similarly, aspects of the ways in which complex numerals interact with nouns are dependent on the preexisting structure of the complex numerals themselves and are extensions or adaptations of the ways in which simple numerals interact with nouns. In turn, and looking ‘further back’, the ways in which simple numerals modify nouns may often reflect structure not involving numerals at all, but attributive adjectives. This is not to say that at each level of development, other aspects of structure may not arise which are sui generis and independent of the prior influences. The facts to be laid out are organized in this way not in a spirit of argument for this broad diachronic hypothesis, but rather in the hope that this organization lends coherence to the whole picture of numeral- noun interaction that will be drawn. Sample. Given the relatively small number of, and the high degree of historical interrelatedness among European languages, it is probably impossible to arrive at a properly unbiased sample. The languages described here are not exhaus- tive of the languages of Europe, but were chosen to give a good geographical spread (North-South, Finnish-Maltese; East-West, Archi-Basque), with no more than three (and usually only one) representatives of any single language group. More information is available about some languages than others; the maximum number of European languages mentioned in connection with some pattern for which statistics are gathered is 16, and the minimum number is 10. The cover- age is therefore not broad by the standards of modern typological surveys, but all the interesting phenomena of numeral-noun interaction that can be found in languages are illustrated and discussed here in some depth. In places, data from non-European languages, especially Hebrew, Sinhala and Shona, are presented by way of contrast with the European data1. Several tables are presented, illustrating statistical trends in the data. In all cases, the skewing of the data evident in these tables is, as far as I can see, also roughly representative of the situation across human languages in general. No phenomena unique to European languages have been discovered, except perhaps for: Plural group numerals in Finnish (and to some extent Icelandic); in these 1.The languages referred to, with the abbreviations used for them, are: Adyghe (Adg), Alba- nian (Alb), Archi (Arc), Avar (Avr), Basque (Bsq), Bulgarian (Blg), English (Eng), Finnish (Fin), French (Fr), Standard German (Grm), Godoberi (Gdb), Greek (Grk), Hebrew (Heb), Hungarian (Hng), Icelandic (Ice), Kabardian (Kbr), Lezgian (Lzg), Maltese (Mlt), Romany (Rmny), Russian (Rus), Scottish Gaelic (ScGl), Shona (Sho), Sinhala (Sin), Welsh (Wls), Zurich¨ German (ZD). 2 languages a morphological marking of the numeral X indicates the meaning ‘X groups of’ (NB not ‘groups of X’). (These phenomena are discussed at some length in subsection 3.3.3.3 of this paper.) Global morphological marking of complex numerals, which is especially thorough in Russian, Finnish and Greek; in such languages, all, or at least many, of the constituent words in a complex numeral, rather than just the last word, are morphologically marked for a particular feature, such as case, gender or ordinality. Perhaps the stimulus of such phenomena being described here will provoke their discovery outside Europe. Certainly, other scholars who have worked on numerals have found these phenomena surprising and unfamiliar. On the other hand, some patterns known from outside Europe may be absent, except in marginal ways, from Europe; for example, no language in this sample has a well developed numeral classifier system. Within the languages surveyed, it is sometimes possible to see a central/peripheral dimension. Thus, apparently unusual features (for European languages at least) are often found in languages spoken at the geographical extremities, such as Finnish, Maltese, Basque, Gaelic, Archi and Kabardian. Such unusual features include: lack of a full range of ordinals (Maltese, see section 2.2), plural group numerals (Finnish, see above, and section 3.3.3.3), switch from plural to singu- lar noun after 10 (Maltese and Scottish Gaelic, see section 5.4.1), non-suppletive ordinal for 1 (Archi, see section 2.2), non-adjectival word order specifically for 1 (Kabardian, see end of section 3.2). Most of the facts reported here were elicited from informants, by the author working with them through a standard questionnaire. Where possible, the infor- mation thus gleaned was checked in grammars. In some cases, the information reported comes only from grammars and scholarly articles. Terminology. It has proved possible to describe most of the data using relatively theory-neutral terms from standard traditional grammar and well-rooted linguistic theory. The most problematic area has been that of agreement, government, and head-modifier relations (see papers in Corbett et al., 1992). In many cases it is clear whether the numeral or the noun is the head, or the modifier, in an attributive numeral-noun construction. But elsewhere it is not so clear, and in order to bring as many languages as possible into comparison with each other, I have simply avoided the issue of whether the numeral, or the noun, agrees with, or governs, 3 the other. What is clear, in all cases, is which expression is a noun and which is a numeral, and it has proved possible to make descriptive statements simply in terms of these syntactic categories. 2 Simple lexical numerals 2.1 Arithmetic range of simple lexical numerals Most of the world’s languages with counting systems have single morphemes for values up to 10, and thereafter use syntactic combinations to express higher num- bers. There are languages with lower bases than 10, usually 5; in such languages, expressions for the numbers 6 - 9 are bimorphemic. I have found no trace of such quinary counting in modern European languages. In Finnish, one can discern the forms for ‘1’ and ‘2’, yhde and kahde in the forms for ‘9’ yhdeksan¨ and ‘8’ kahdeksan respectively , indicating a bimorphemic subtractive origin of these numerals, but modern Finns are typically not conscious of the forms for ‘8’ and ‘9’ being bimorphemic, and can assign no meaning to *ksan. Throughout Europe generally, with few exceptions, the onset of complex syntactic numerals comes at ‘11’. See section 4.3 for further discussion. 2.2 Distinct counting forms The numeral which is used to quantify a noun in a noun phrase is not necessarily the same form as the corresponding numeral in the conventional recited counting sequence. I refer here to differences other than those, such as inflections, due to the influence of the sister noun or mother noun phrase. There are various de- grees of idiosyncratic difference between a counting numeral and a quantifying (attributive) numeral. The most extreme difference is where the counting numeral is a suppletive variant of the quantifying numeral (or vice versa). For example, when counting in Maltese, 2 is expressed as tnejn, but in construction with a noun, the word for 2 is usually ze˙ wg˙; expressions such as tnejn kafe ‘two coffees’, complicate the picture. Russian has a special counting numeral for ‘1’ raz, whereas the corresponding attributive form is some suitably inflected form of odin. The best known example of this sort occurs outside Europe, in Chinese, where there are also two quite different words for 2, depending on whether one is reciting the counting sequence or expressing a proposition about some collection of two objects. The second 4 word in the standard counting sequence is er` , whereas the word meaning 2 used with nouns (and their accompanying classifiers) is liangˇ . Less extremely idiosyncratic are cases where a counting numeral is phono- logically similar to the quantifying form, but not predictable from it by a rule applying to other forms. For example, in German, the counting numeral eins, ‘1’, is similar to, but not precisely predictable from any of the quantifying forms ein, eine, einen, eines, einer, einem. Sometimes the counting numeral is the same as a pro-form numeral; German eins happens to be identical to one of the available pro-forms for an indefinite sin- gular noun phrase. Similarly in Hungarian, 2 is either ket´ or ketto¨. The longer form is used in counting and when no quantified noun is present (i.e. as a pro-form for an indefinite noun phrase); the shorter form is used to quantify a noun. ket/k´ etto¨ is the only simple lexical numeral in Hungarian which shows such a difference.